Blockchain-Based Transformer-Assisted Multi-Agent Reinforcement Learning for Resource Allocation and Computation Offloading in 5G Private Networks

The rapid development of 5G private networks has brought demands for ultra-low latency, high bandwidth, and high reliability. However, resource allocation and task scheduling in multi-task and multi-user scenarios still pose significant challenges. To address this, this paper proposes a blockchain-based and Transformer-assisted multi-agent reinforcement learning algorithm for resource allocation and computation offloading optimization in 5G private networks. By constructing a total cost function that comprehensively considers task completion time, energy consumption, and service quality, and employing a multi-agent reinforcement learning method, each device can adaptively adjust its strategy. The Transformer model is introduced to effectively handle high-dimensional state information through its self-attention mechanism, enhancing the expressiveness and learning efficiency of the policy network. Combined with blockchain technology, the proposed method reduces the cost of blockchain operations while ensuring system security. Experimental results show that the proposed algorithm outperforms baselines in terms of total cost, task completion time, energy consumption, and QoS satisfaction rate, validating its effectiveness and superiority.